We reexamine prior studies' conclusion that accruals are less persistent than cash, focusing on two aspects of persistence that are crucial to determining its properties. The first (time specificity) refers to the fact that persistence describes how current-period shocks to income translate into next-period income. Traditional measures of accruals are, however, functions of current- and non-current-period transactions. We show that the inclusion of non-current-period transactions leads to a downward (upward) bias on the persistence of accruals (cash flows). We develop alternative measures of accruals and cash flows that are not misaligned and show that the differential persistence of cash flows over accruals is more than 70% smaller using these measures. The second aspect of persistence is firm-specificity. Specifically, we evaluate persistence using firm-specific estimations and find that more than 85% of firms show no evidence that accruals are less persistent than cash flows.

We contrast the informativeness of earnings and dividends for firms with dual class and single class ownership structures. Results of both across-sample tests (which explicitly control for factors influencing ownership structure and informativeness) and within-sample tests (which implicitly control for factors associated with ownership structure) show that earnings are generally less informative, and dividends are at least as (if not more) informative, for dual class firms. We interpret these results as suggesting that the net effect of dual class structures is to reduce the credibility of earnings and to enhance the salience of dividends as measures of performance.

Bayesian learning implies decreasing weights on prior beliefs and increasing weights on the accuracy of the analyst's past forecast record, as the number of forecast errors comprising her forecast record (its length) increases. Consistent with this model of investor learning, empirical tests show that investors’ reactions to forecast news are increasing in the product of the accuracy and length of analysts’ forecast records. Moreover, the Bayesian learning predicted by our model is more descriptive of investor reactions than is a static model which predicts that investors’ responses condition only on the prior accuracy of the analyst.

We investigate whether investors price accruals quality, our proxy for the information risk associated with earnings. Measuring accruals quality (AQ) as the standard deviation of residuals from regressions relating current accruals to cash flows, we find that poorer AQ is associated with larger costs of debt and equity. This result is consistent across several alternative specifications of the AQ metric. We also distinguish between accruals quality driven by economic fundamentals (innate AQ) versus management choices (discretionary AQ). Both components have significant cost of capital effects, but innate AQ effects are significantly larger than discretionary AQ effects.